flatness measurement datum
Photo chemical machining for stainless steel shim is a flat process, and that is exactly why flatness is the property customers complain about. Etching itself applies no force, so a freshly etched shim starts out flatter than a stamped one. Then it is rinsed, dried, inspected, packed, shipped, maybe plated, and handled — and every one of those steps can put a bow into a part that weighs a few grams. A flatness number is also the one number a customer can check with a ruler. The useful conversation is not “why is it not flat” The useful conversation is not “why is it not flat” but “flat under what condition, measured how, and kept flat by what”. A shim that arrives flat and bows in the customer’s tray is still a faulty shim, and the cost of a bow is not the part but the sorting and the re-quote; flatness is also the property a drawing is least likely to state properly.
1. Flatness Is the First Complaint
Ask a shim buyer what went wrong last year and the answer is usually flatness rather than dimension.
Three reasons. A bow is visible and easy to measure, so it is the first defect anyone notices. It matters functionally, because a shim that is not flat does not seat and does not set the gap it was designed for. And it is frustrating, because it often appears after the parts left the supplier in good condition.
That last point is the one worth designing around: flatness is a property that has to survive the supply chain, not just the etch.
2. Why a Thin Part Bows
A thin sheet bows for one of three reasons, and they are worth separating. A sheet that is flat on arrival can still be stressed, and the etch is what reveals it; a strip that has been slit carries more stress at the slit edge, which is why the same design can bow differently from a wider sheet, and a part that curls consistently in one direction is telling you which face lost metal.
The first is residual stress in the incoming strip. Rolled strip carries stress, and when one face is removed — by etching, by polishing, by blasting — the balance is broken and the part curls.
The second is an unbalanced etch. If one face is etched more than the other, or if a large area is removed from one side and not the other, the part curls toward the side with less metal.
The third is handling. A part that is thin enough to be flexible is thin enough to be bent by a hand, a tweezer, a vacuum pickup or a bag that is too tight.
Only the third is really a “flatness” problem in the metrology sense; the first two are a stress problem that shows up as flatness.
3. What the Etch Does
Etching removes metal from both faces, which makes it the most symmetrical of the common processes — and it still introduces bow.
Three mechanisms. A two-sided etch that is not equal removes more from one face, which bows the part. A pattern with a large open area on one side and a mostly solid area on the other creates an unbalanced section. A balanced design and a balanced etch are two different things and both are needed. And the resist itself applies a small stress And the resist itself applies a small stress while it cures and while it is stripped. A pattern that leaves a large solid border on one face and a large window on the other is the classic unbalanced section, and an etch that is symmetric in area is not always symmetric in stress because the resist is on both faces.
The practical consequence is a design rule: keep the metal distribution as balanced as the function allows, and where it cannot be balanced, expect to control flatness by handling and packing rather than by the etch alone.
4. Free Versus Constrained
This is the specification trap, and it causes more disputes than any manufacturing issue.
A shim measured free on a surface plate will show a bow. The same shim clamped, or lying under its own weight in a fixture, may show almost none. Both are legitimate measurements and they answer different questions.
A free-state number answers “will this part be flat when someone picks it up”, which is what a customer feels. A constrained number answers “will this part function once installed”, which is what the stack cares about.
A drawing that says “flatness 0.05” without saying which state is a drawing that will be argued about. A free-state limit and a constrained limit can differ by an order of magnitude and both can be honest; the state is not a detail, it changes the number by more than the tolerance, and a supplier quoting free state while a customer measures constrained will disagree forever. A drawing should say free state unless the function demands otherwise, and then it should say the fixture. It should name the state, the support (three points, a plate, a fixture), and the measurement method.
5. How to Measure It
Three measurements cover almost every case, and they should be named on the drawing.
A surface plate and a height gauge or a dial indicator, with the part free, gives a bow number that correlates with what a customer sees.
A three-point support with the part free gives a number that removes the effect of an uneven plate.
A constrained measurement, in the fixture the part will actually live in, gives the functional number.
The three do not agree, and they are not supposed to. The mistake is to pick one without saying which. A height gauge on a surface plate is crude and it is also the measurement that matches the complaint, a dial indicator on granite is worth more than a CMM when the question is how the part sits, and a measurement taken with the part held in a hand is a measurement of the hand.
6. Stress, Temper and the Material Choice
The material choice moves flatness as much as the process does.
A harder temper carries more residual stress, so a full-hard shim bows more than an annealed one at the same thickness. A material certificate does not state residual stress, so the condition has to be specified by name. A grade that is stress-relieved before etching bows less.
A very thin shim in a hard temper is the worst combination for flatness, and it is also the combination most often specified for spring function — which is why the conflict has to be resolved at design rather than at inspection.
Where flatness is critical, an annealed or stress-relieved condition is worth more than a tighter dimensional tolerance. A stress-relieved strip costs more per kilogram and less per rejected lot, and a hard temper specified for spring function is a design decision with a flatness cost attached.
7. Packing and Handling
Flatness that survives to the customer is a packing problem as much as a process one.
Three practices matter. Parts should travel flat and supported, not loose in a bag where they can bend against each other. Interleaving with a stiff card or a foam sheet prevents the stack from taking the shape of the bag. A crate that is dropped is a crate whose parts are bent, however well they were packed. And the packing should be rigid enough that the box, not the parts, absorbs the handling.
Cleanliness and flatness interact here too: a part that is handled with bare hands for counting gets bent at the edge, so counting by weight or by a counted stack is a flatness control as well as a labour saving. A bag filled to the top is a bag that has bent the parts at the bottom, and a flat pak with a card insert is cheap and is the difference between a good part and a return.
8. What the Drawing Should Say
A drawing for a thin etched shim should state the flatness limit, the state it applies to (free or constrained), the support condition, and the measurement method.
It should also state the material condition, because a hard temper and an annealed one are different flatness products at the same tolerance.
And it should say what happens on a bow: a re-flatten, a concession, or a reject — because a part that could be flattened by a simple operation should not be scrapped for a number.
The recurring mistakes are a flatness callout with no state, A disposition clause on flatness is worth more than a tighter number because it tells the supplier what to do instead of scrapping, and a note that says flat within reason is not a specification because it will be read as the supplier’s reason. The recurring mistakes are a flatness callout with no state, a hard temper specified for a part that also needs to be flat, parts shipped loose in a bag, and a measurement taken on a plate that is itself not flat.
